Understanding the Energy Efficiency Operational Indicator...2 Understanding the Energy Efficiency...

1

Understanding the Energy Efficiency Operational Indicator:

An empirical analysis of ships from the Royal Belgian Shipowners’ Association

Final Report and Appendices May 2015

2

Understanding the Energy Efficiency Operational Indicator: an empirical analysis

of ships from the Belgian Shipowners’ Association

Sophia Parker, Carlo Raucci, Tristan Smith, and Ludovic Laffineur

UCL Energy Institute, Royal Belgian Shipowners’ Association

Abstract

This paper explores the determinants of the Energy Efficiency Operational Indicator

(EEOI1) by decomposing the indicator into sub-indices that reflect the technical

characteristics and logistics of ships. We examine these sub-indices by first constructing a

model that mathematically describes the components that comprise the EEOI. A panel

dataset of ships’ fuel consumption parameters and transport work is used to estimate

these sub-indices. We find that there is a relationship between technical efficiency

(EEDI) and EEOI across different ship sizes, but there is a wide dispersion of EEOI

values within a ship size class. This can be explained by the variation in logistics factors,

with little evidence of correlation between EEOI and any one logistics factor. For all of

the types and sizes considered, variations in EEOI can be explained only by considering

contributions from a combination of the logistics factors.

Keywords: Energy Efficiency Operational Indicator (EEOI), shipping, MRV,

energy efficiency, carbon intensity

1. Introduction

Shipping is commonly cited as the most efficient transport mode. When expressed as a

generalization (across all ship types), this is rarely disputed. However, recent discussions

and attempts to quantify the more detailed energy efficiency characteristics of the

existing ship fleet have been met with criticism. For example, among the objections to

previous analyses, studies have had issues related to unrepresentative input data, limited

real-world operational data to reflect actual operational conditions, and incomplete

quantification of technical versus operational efficiency characteristics. Many of these

objections are well-founded, due to the generally poor quality of data describing the

existing fleet of ships and the wide-ranging parameters that influence the performance

and therefore efficiency of ships in their day-to-day operation (as opposed to an artificial

‘calm’ sea or acceptance trial).

Increasing the motivation for a more comprehensive analysis of energy efficiency is the

ongoing debate about how shipping’s air pollution and greenhouse gas emissions should

1 The EEOI is the total carbon emissions in a given time period per unit of revenue tonne-miles. A lower EEOI means a ship is more energy efficient in its operations.

3

be regulated. In January 2013, the EEDI (Energy Efficiency Design Index) came into

force, requiring all newbuild ships to meet a minimum energy efficiency standard. In the

same regulation annex, the SEEMP (Ship Energy Efficiency Management Plan)

recommends the use of the EEOI (Energy Efficiency Operational Indicator) as a

measurement of the energy efficiency of existing ships.

The EEOI, or annual fuel consumption divided by transport work, can be considered as

the annual average carbon intensity of a ship in its real operating condition, taking into

account actual speeds, draughts, capacity utilization, distance travelled, and the effects of

hull and machinery deterioration and weather. Although the EEOI is referred to as an

indicator of energy efficiency, it is technically more accurate to refer to the EEOI as a

measure of carbon intensity as the units are in gCO2/t.nm. The US energy efficiency

indicator is measured in joules/hour and therefore is more defensibly energy efficiency

because the numerator is measured in joules of energy. Despite these differences, if the

fuels are similar in carbon content then the CO2 and joules should be consistent so that

carbon intensity is a proxy for energy efficiency. On the other hand, a ship that consumes

LNG would differ in energy content compared to one using HFO and therefore a

correction would need to be applied for the relative carbon and energy intensities of the

fuel before their energy efficiency can be compared. For the rest of the report, we will

refer to the EEOI as an energy efficiency indicator, except in Section 6, when the merits

of carbon intensity are discussed in light of alternative indicators proposed at the IMO.

The EEOI and the data and methods associated with it were originally conceived for

policy purposes, evidenced by the European Commission’s proposal to require ships

exceeding 5,000 GT to monitor and report their operational energy efficiency starting in

2018 on all voyages to, from, and between EU ports. In light of this potential legislation,

ship owners and their associations are trying to better understand the drivers of the

proposed Energy Efficiency Operational Indicator (EEOI) in order to prepare themselves

for future environmental regulation. Discussions at the IMO indicate that such MRV

initiatives could serve as initial phases toward eventual in-use ship fleet efficiency

standards.

The collection of fuel consumption data, as required by the MRV policy, has lead to

speculation about how the EEOI could be extended to other regulations or for commercial

purposes. One of the commonly cited barriers in the shipping industry is the lack of

sufficient information on the technical efficiency of a ship operated in real operating

conditions when a ship is chartered (Rojon & Smith, 2014). Although there is publicly

available data that approximates the technical efficiency of a ship when it is built, the

efficiency of a ship in its designed condition at age 0 does not necessarily equal the ship’s

technical efficiency because the formula (EEDI) makes assumptions2 about parameters

2 The EEDI is measured the design speed, and the specific fuel consumption

4

that determine efficiency. Furthermore, as a ship ages, the specific parameters that

determine its fuel consumption change over time due to a gradual deterioration of the

hull’s surface and fouling due to marine growth. For example, two ships which appear

identical in their design characteristics can perform differently due to a difference in

maintenance or retrofitting which would not be observed in the EEDI.

Although ship owners measure fuel consumption and cargo information, it is not known

to what extent this data is used for improving operations. As a result, the industry lacks a

detailed understanding of the consequence of energy efficiency interventions on its

emissions (e.g. slow steaming). And more broadly, bottom-up estimates of shipping

emissions (e.g., those used by the IMO and other groups) can potentially lack credibility

or sources of validation.

Industry groups have sought to address these failings, including work undertaken already

by the RBSA. In 2009, the RBSA collaborated with the Flemish Institute for

Technological Research (VITO) to create a study on the Energy Efficiency Operational

Indicator. The purpose of the study was to identify the gaps in the interim guidelines

(MEPC/Circ. 471) developed within the IMO to determine the EEOI. The index was

tested on 41 ships under the Belgian flag. The results of the study were presented to the

IMO (GHG-WG 2/3/1) through the Belgian maritime administration in 2009. As the

EEOI is an aggregate number, it is difficult to disentangle the influence of these

confounding factors. Therefore a database was established in order to determine the

contribution of factors such as ballast voyages and port time to the index due for each

individual ship. The main message of the study was that breaking down the basic formula

leads to better transparency of the causes of variation of the EEOI and may help to

improve operational and environmental performance for ship operators.

This paper further studies the drivers of the indicator by carrying out a series of analyses

on a set of owner-reported data, similar to the data that will be used to comply with the

future legislation. As well as calculating the carbon emissions and values of EEOI for

ships in the RBSA’s owner’s fleets, the study will decompose the EEOI into sub-indices

(technical and logistics factors) and in terms of the contribution of the laden, ballast and

port segments to EEOI for 94 ships in the bulk carrier, chemical tanker, container,

liquefied petroleum gas and oil tanker sectors over the period 2008-2014, in which there

was variation in market factors such as fuel prices and freight rates. These market factors

have influenced the way in which the ships were operated, including the speed and the

employment opportunities available and undoubtedly has cascaded into changes in the

EEOI over time. A number of other operational energy efficiency indicators have been

proposed by member states of the IMO. In light of these other efficiency indicators, the

analysis will also compare the EEOI to alternative energy efficiency indicators. The

experience of computing the EEOI has also not been well documented. This paper will

5

shed light on the process and challenges of calculating the EEOI, as well as the

uncertainty in the estimates calculated.

2. Expressions for technical and operational efficiency and the interconnection of the

two types of efficiency

This section discusses the formulation of indices for efficiency, including a suggested

additional indicator that represents the technical efficiency of a ship at a given point in

time. A full derivation of the equations used is contained in Appendix A.1.

The annual EEOI, or annual total carbon emissions divided by transport work, can be

considered as the annual average efficiency of a ship in its real operating condition,

taking into account actual speeds, draughts, capacity utilization, distance travelled, and

the effects of hull and machinery deterioration and weather.

A metric used for quantifying the operational efficiency of shipping is the EEOI (IMO

MEPC.1/Circ.684, 2009):

(1)

𝐸𝐸𝑂𝐼 =∑ ∑ 𝐹𝑖𝑗𝐶𝑗

𝐹𝑗𝑖

∑ 𝑚𝑖𝐿

𝑖 𝐷𝑖𝐿

where:

i = the voyage

j = the fuel type

𝐹𝑖𝑗= the amount of fuel consumed for the voyage i and fuel type j

𝐶𝑗𝐹= the carbon factor for fuel type j

𝑚𝑖𝐿= cargo mass of voyage i

𝐷𝑖𝐿= distance travelled in loaded voyage i

Analogous to the use of the EEOI to estimate the operational efficiency of a ship, the

technical efficiency of a ship or energy efficiency technical indicator (EETI) can be

defined as the energy efficiency (gCO2/tonne-nautical mile) of a ship in a reference

operating condition (speed and draught):

(2)

𝐸𝐸𝑇𝐼 =𝐶𝐹𝐹𝑑

𝑟𝑒𝑓

𝑣𝑟𝑒𝑓𝐷𝑊𝑇24

where:

𝐶𝐹 = the average carbon factor of the fuel used

6

𝐹𝑑𝑟𝑒𝑓

= the daily fuel consumption at a reference speed and draught

𝑣𝑟𝑒𝑓= the reference speed3

𝐷𝑊𝑇 = the deadweight tonnage of a ship

The EETI is a ship’s estimated technical efficiency in real operating conditions at a

specific point in time, whereas the EEDI (gCO2/t-nm), is the ship’s design technical

efficiency at the start of its life and under specific EEDI assessment conditions.

Differences between a ship’s EEDI and EETI could arise due to fouling, modification of

technical specifications (such as retrofitting) or because the EEDI trial performance

cannot be recreated in real operating conditions.

The EEOI and the EETI, when estimated from measurements of a ship’s daily fuel

consumption and activity, are inextricably linked because they have overlaps in their

input data. The mathematical derivation of the two formulae can be used to show how

EEOI can be decomposed into a number of technical and logistics factors, one of which is

the ship’s EETI. This is useful as a means to break EEOI down into a series of drivers

that each influence the overall value. As the only ‘technical factor’, the EETI indicates

the technical efficiency contribution to the EEOI quantification, while three logistics

factors indicate the contribution of the specifics determining the ship’s commercial

operation (speed and utilization).

(3)

𝐸𝐸𝑇𝐼~𝐸𝐸𝑂𝐼𝑚𝑡

𝐿

𝐷𝑊𝑇

𝑑𝑡𝐿

(𝑑𝑡𝐿 + 𝑑𝑡

𝐵)

𝑣ℎ𝑡𝐿

((∑ (𝑣ℎ𝑡

𝑜𝑝,𝑖

𝑣𝑟𝑒𝑓)

𝑛

(𝑇𝑜𝑝,𝑖

𝑇𝑟𝑒𝑓)2/3𝑝𝑖=1 ) 𝑝⁄ ) 𝑣𝑟𝑒𝑓

where:

𝑇𝑜𝑝,𝑖 = the operating draught at passage or passage segment i

𝑇𝑟𝑒𝑓 = the reference draught

𝑣ℎ𝑡𝑜𝑝,𝑖

= the average operating speed for a passage or passage segment i (nautical

miles/hour).

p = total number of passages or passage segments

𝑑𝑡𝐿= the days a ship is sailing loaded during period t

𝑑𝑡𝐵= the days a ship is sailing ballast during period t

𝑑𝑡𝑃= the days a ship is in port during period t

𝑚𝑡𝐿 = the average cargo mass during period t

𝑣ℎ𝑡𝑙 = the average loaded speed per hour h in time period t

3 The reference speed may or may not equal the design speed.

7

Equations (3) shows that it is possible to formulate EEOI or EETI in terms of one another

if a number of extra details about speed, cargo and the loaded/ballast voyages are known.

These are, as represented in the right hand side of (3):

𝑚𝑡𝐿

𝐷𝑊𝑇= the average payload utilization

𝑑𝑡

𝐿

(𝑑𝑡𝐿+𝑑𝑡

𝐵)= the allocative utilization or ratio of laden days to total operating days

𝑣ℎ𝑡𝐿

((∑ (𝑣

ℎ𝑡𝑜𝑝,𝑖

𝑣𝑟𝑒𝑓)

𝑛

(𝑇𝑜𝑝,𝑖

𝑇𝑟𝑒𝑓)2/3𝑝𝑖=1

) 𝑝⁄ )𝑣𝑟𝑒𝑓

= the speed and draught factor

The average payload utilization is always less than 1, the allocative utilization is also

always less than one, and the speed factor could be greater than, equal to, or less than 1.

Although commonly, especially recently, it will be greater than 1 – demonstrating slow

steaming.

In operation, a ship with a higher EETI value can offset this lower efficiency

disadvantage by obtaining a higher average payload utilization, allocative utilization,

speed factor, or some combination of these factors. As expected, this shows that the EEOI

is highly influenced by how a ship is commercially operated, and only partially

influenced by the technical efficiency of the ship. The derivation in this section of EETI

and the connection between EEOI and EETI is used both to illustrate this point, and to

introduce the concept of EETI which is calculated explicitly using the data in this study,

with results presented in Section 6.

3. Description of the data

3.1 Ships covered in the dataset Data for this paper comes from five companies who are members of the Royal Belgian

Shipowners’ Association (RBSA). RBSA has provided spreadsheets of data supplied by

the ship owners. Where possible, other documentation such as noon report or arrival and

departure report has also be provided by RBSA.

Table 1 describes the types of ships for which we are able to analyze the data, as for some

ships the data provided is not consistent or complete so some ships had to be excluded. A

filter has been applied that excluded every year of a ship for which more than 30% of

voyages has not complete information. The majority of the ships analyzed are bulk

carriers, accounting for 42% of the sample, followed by oil tankers (25%) and liquefied

petroleum gas carriers (23%).

8

Table 1 Ships sample

Ship type/size IMO size range

IMO

size

categ

ory

Original

number

of ships

Ships

analyzed

% ships

excluded

Bulk carrier (dwt) 58 43 26%

(10000-34999) 2 15

(35000-59999) 3 4

(60000-99999) 4 1

(100000-199999) 5 20

(200000-+) 6 3

Chemical tanker (dwt) 2 2 0%

(10000-19999) 3 2

Container (TEU) 10 10 0%

(1000-1999) 2 3

(2000-2999) 3 7

Liquefied petroleum

gas (cbm) 24 17 29%

(0-49999) 1 15

(50000-199999) 2 2

Oil tanker (dwt) 30 22 27%

(120000-199999) 7 17

(200000-+) 8 5

Grand Total 94

The data is an unbalanced panel of individual ships over time because not all ships are

reported in each year over the period 2008-2014.

Table 2 shows the representation of ships by year in the sample. There is data for bulk

carriers in years 2008-2013, while data was not available for several years for each of the

other ship types in the dataset.

9

Table 2 Number of ships in the data sample by year

Ship type 2008 2009 2010 2011 2012 2013 2014

Bulk carrier 11 13 21 33 41 37 0

Chemical tanker 0 0 0 2 2 0 0

Container 0 0 6 9 8 9 0

LPG 4 10 12 12 4 0 0

Oil tanker 0 0 4 12 7 9 13

In order to obtain additional information on each ship’s technical specifications

(deadweight, age, installed power, Specific Fuel Oil Consumption and reference speed)

data was taken from Clarkson Research Services, by matching the IMO number provided

by each company’s data. This information allowed us to estimate the ships’ Energy

Efficiency Design Index (EEDI) using the formula provided in Germanischer Lloyd SE

(2013). The estimate is similar to an EIV (Estimated Index Value) used in the calculation

of EEDI baselines, but uses SFOC as reported in Clarksons. It is an estimate because

there is no validation of the calculation’s input data.

Table 3 provides the summary statistics of the average technical specifications. There is a

strong relationship between the technical efficiency of a ship, represented by the EEDI,

and the size of the ship (in deadweight tonnes or DWT). For all ship types, EEDI is

decreasing with size, meaning that the energy efficiency is higher.

Table 3 Average technical specifications. Source: Clarkson Research

Ship type

/IMO size

Size range No. of

ships Mean

age Mean

DWT Mean

design

speed

Mean

EEDI

Bulk carrier (dwt) 43 6 109,372 14.48 5.44

2 (10000-34999) 15 4 33368.93 14.03 9.15

3 (35000-59999) 4 7 54997.5 14.73 5.53

4 (60000-99999) 1 8 76588 14.40 3.98

5 (100000-199999) 20 8 176760.7 14.78 3.11

6 (200000-+) 3 3 205143.3 - -

10

Chemical

tanker (dwt)

2 13 14582.5 13 16.309

3 (10000-19999) 2 13 14582.5 13.00 16.31

Container (TEU) 10 8 25141.8 20.52 14.88

2 (1000-1999) 3 10 16701.67 19.53 16.37

3 (2000-2999) 7 6 33582 21.50 13.39

LPG (cbm) 17 15 35952.8 15.53 12.66

1 (0-49999) 15 10 12718 14.82 18.11

2 (50000-199999) 2 21 59187.5 16.25 7.21

Oil tanker (dwt) 22 11 230798 15.48 2.73

7 (120000-199999) 17 12 154202.8 14.98 3.29

8 (200000-+) 5 10 307393.6 15.98 2.16

Grand Total 94 11 82,169 16 10

3.2 Data used to estimate the EEOI and subindices

We calculate the annual EEOI using detailed data on the fuel consumption of a ship and

revenue tonne-miles per sea passage. As also identified in the VITO study (VITO, 2009),

the format for reporting this detail varies by company; a sea passage could be defined as

starting from one port and ending in another port or starting at sea and ending at sea. In

some cases, the sea passage is not specified. In this case, we derive the passage from two

temporal consecutive records.

While each company has its own internal procedure and format for collecting this data,

each company collects data on fuel consumption separately from cargo information.

Therefore, the data had to be merged together. An overview of the procedure and data

checking is described in Appendix A.2. Each company’s data was checked for

consistency of the parameters required for the calculation (distance sailed, speed, hours of

operation) and processed to obtain a standardized dataset that was uniform for all ships.

This involved extensive filtering, cleaning and checking of the data.

There are a number of missing fields in the unprocessed data. We therefore estimate

speed, distance or hours when at least two of the variables are provided. When all

variables are known, we verified the hours of service using the distance/speed

relationship to ensure that the triple is consistent. We also checked for outliers, described

in Appendix A.2.

11

There are a number of cases in which we could not calculate the EEOI for a specific

voyage. For example, if the fuel consumption per laden voyage is known, but the distance

for these voyages is missing and speed is nonzero, then we excluded this voyage from the

analysis to avoid overestimating the numerator without a corresponding tonne-miles

statistic. These exclusion cases are described in Appendix A.2. Finally, the annual EEOI

was calculated by aggregating the sea passages for which we have both valid fuel

consumption and revenue tonne-miles data.

4. EEOI results

4.1 Aggregated annual EEOI for all ships (grouped by ship type and size)

We calculated annual EEOI for all ships in the database for each year (2008-2012) for

which there was valid data. Figure 1 shows an overview of the annual EEOI in relation

with the DWT for each ship type4.

Figure 1 Annual EEOI and DWT grouped by ship type

The variation in annual EEOI for each ship type varies. For example, for large bulk

carriers the annual EEOI varies between 5 and 20 gCO2/t.km, while the variation in the

EEOI for smaller bulk carriers is wider, between 5 and 40 gCO2/t.km. The results are

shown in Figure 1 for each ship grouped by ship type. Note that the same ship will appear

multiple times if data is available on the ship for multiple years. It can be seen that the

12

relationship between size (measured in DWT) and operational efficiency is not obvious

given the wide dispersion of values for ships of a similar size (a higher EEOI means the

ship is less efficient).

Figure 1 also shows there are several notable outliers. Generally, high EEOI values are

often the result of a very low allocative utilization5 and payload utilization. For example,

the highest annual EEOI value (1428 gCO2/t.km) was from the LPG ship type, which for

the size class of 50,000-199,999, had an allocative utilization of about 5% and a payload

utilization is about 22% in 2010. This is low compared to even the lower size class of

LPG; in 2010, the allocative utilization for the 0-50,000 class size was 47% and had a

payload utilization rate of 41%. The outlier values are affected by the availability of the

data in that year. For example, the data with the highest EEOI values is only available for

about 29 days in the year.

As the technical efficiency (EEDI) improves with size due to economies of scale, we also

present the EEOI for each ship type and size (by IMO size categorization). Figure 2

shows the annual EEOI by DWT grouped by size for the bulk carriers in the sample. A

non-linear curve fitted to the data shows that when outliers are excluded, DWT explains

46% of the total variation in the EEOI, as defined by the coefficient of determination or

R2. The R

2 produced from the model 𝐸𝐸𝑂𝐼 = 815𝐷𝑊𝑇−.39 + 𝜖 is an estimate of how

much variation in EEOI is explained by DWT in the data sample and can be used as a

measure of the goodness of fit to the data. A higher R2 indicates a better fit, with 100%

representing the regression line that perfectly fits the data. As predicted, DWT does not

explain even half of the variation in the EEOI. This highlights the importance of

examining the operational or logistics factors driving variation in EEOI values.

5 Allocative utilization is the ratio of days laden to total sailing days.

13

Figure 2 Annual EEOI and DWT for Bulk carrier by size

Figure 3 shows the distribution of annual EEOI for bulk carriers by size class between

2008-2013. This data was the most complete of the ship types in the study and will

therefore be highlighted more in the report to explain the drivers of EEOI. The figure

shows that the EEOI is monotonically decreasing in size, ranging between 5.74 for the

largest size class (200,000+ DWT) and 13.42 (10,000-34,999) gCO2/t.nm. Although size

clearly does influence the EEOI values, the boxplots6 show that there is variation within

each size class. Section 6 will decompose the EEOI into sub-indices in order to explain

the factors driving variation in these values.

6 On each boxplot, the central mark is the median, the edges of the box are the 25th and 75th percentiles, the whiskers extend to the most extreme data points not considered outliers, and outliers are plotted individually. For example, the median value for carriers of 200,000+ DWT is 5.74. The mean is also plotted as a green diamond.

14

Figure 3 Annual EEOI bulk carrier by size

4.2 Variants of annual EEOI

Variants of the annual EEOI are also calculated for all ships in the database and all years,

on a ‘per-ship’ basis. These variants include:

The contribution of loaded, ballast and port EEOI per voyage

The laden voyage EEOI presented as a rolling average alongside the annual EEOI

Figure 4 shows the annual, voyage, and rolling average EEOI for a bulk carrier ship of

size 5 (100,000-199,999). Appendix A.5 shows the same type of plot for all ships. The

laden and ballast EEOI are of similar magnitude, with the exception of 2012, when the

ship incurred a higher laden EEOI than ballast EEOI, reaching a value of nearly 6

gCO2/t.nm. This is driven by a lower payload utilization of about 42%.

Section 2 showed that the EEOI can be broken into sea EEOI, consisting of the fuel

consumption when a ship is sailing divided by transport work, and port EEOI, which is

the fuel consumption when a ship is in port divided by transport work. The data shows

that port EEOI represents a significantly smaller share, only accounting for 7% of EEOI

because a ship is relatively stationary when in port and thus consumes a small proportion

of total annual fuel consumption.

15

Figure 4 Variations of EEOI, voyage EEOI and rolling average for bulk carrier ship

Also presented in Figure 4 is the laden EEOI of individual voyages, which measures the

fuel consumption on an individual voyage divided by the transport work performed for

that voyage. The rolling average takes the average of 3 consecutive voyages. These

numbers can be volatile, especially notable in 2012. The points which show high

increases in the laden EEOI can be explained by a low cargo value compared to other

observations. In particular, for the highest value of laden EEOI in 2012 (5.91 gCO2/t.km),

the payload utilization for that specific voyage is about 42%, a low value compared to an

average of 91%. In addition, there are more laden voyages in 2012 which explains why

the annual EEOI is lower compared to the other years.

4.3 Emissions, distance, service hours

We calculated the emissions, fuel consumption by type, distance travelled, and service

hours per ship. Figure 5 illustrates the CO2 emissions, fuel consumption, distance

travelled, and hours of service for the same bulk carrier represented in Figure 4.

Appendix B provides a similar plot for all ships. This ship has sufficient data coverage, as

seen by the minimal “out” area or data that was excluded from the EEOI calculation. The

average annual proportion of emissions in laden is 49%, compared to 44% for ballast, and

7% in port. HFO contributes the most to fuel consumption compared to MDO, averaging

annually 94% over 2008-2013 period.

The hours spent in port are much higher than the emissions in port because of the low

amount of fuel consumption burned in port due to their relatively stationary position. A

16

similar result can be seen for the other ships in the sample7, which justifies our focus on

EEOI at sea as the main contributor to total carbon emissions.

Figure 5 Emissions, fuel consumptions, distance travelled and hours of services for ship

ID 81

7 The annual average proportion of port emissions for ships in the sample is about 9%

17

4.5 Tabular results per year for all ship types and sizes

Table 4 Results year 2008

Type Size Size

sample

Mean dwt Mean

days at

sea

Mean at

sea speed

EEOI (gCO2/t.nm) Median

allocative

utilization

Median

payload

utilization

Mean

transport

work per

ship

(tonnes) (knots) Median Lower

quartile

Upper

quartile

(%) (%) billion t.nm

Bulk carrier 35000-

59999

1 53505 118 13.75 11.66 11.66 11.66 57.4 86.4 1.04

Bulk carrier 60000-

99999

1 76588 98 19.30 8.06 8.06 8.06 43.4 87.6

1.32

Bulk carrier 100000-

199999

9 174843 118 14.21 8.30 7.25 9.86 41.0 95.4

2.68

LPG 0-49999 4 18135 112 14.63 65.41 44.20 107.76 41.9 62.9 0.19

18


Type Size Size sample Mean dwt Mean

days at

sea

Mean at

sea speed


allocative

utilization

Median

payload

utilization

Mean

transport

work per

ship


quartile

Upper

quartile

(%) (%) billion

t.nm

Bulk carrier 35000-

59999

2 53459 75 13.65 11.644 10.9804 12.3068 58.16 92.13 .70


199999

11 175625 89 14.02 7.50 6.58 8.11 49.00 97.50 2.28

LPG 0-49999 10 11910 91 13.95 115.58 63.98 144.38 46.44 43.81 0.08

19


Type Size Sample

size

Mean dwt Mean

days at

sea

Mean at

sea speed


allocative

utilization

Median

payload

utilization

Mean

transport

work

per ship


quartile

Upper

quartile

(%) (%) billion t.nm

Bulk carrier 10000-

34999

3 33353 44 13.45 14.36 13.50 18.04 71.6 93.2 0.3

Bulk carrier 35000-

59999

4 54998 86 13.71 13.36 10.59 14.85 55.3 91.3 0.8


199999

13 175649 112 14.22 7.65 7.12 8.59 47.3 93.1 2.6

Bulk carrier 200000-+ 1 205097 88 14.36 7.12 7.12 7.12 49.5 86.6 2.7

Container 2000-2999 6 33607 235 17.47 30.22 27.04 33.73 100.0 70.0 2.3

LPG

0-49999 11 9157 115 13.65 146.57 96.42 157.40 47.0 40.6 0.1

LPG

50000-

199999

1 64220 29 12.94 1428.24 1428.24 1428.24 5.8 22.7 0.0

Oil tanker 120000-

199999

3 154784 81 11.22 11.99 10.32 16.71 52.6 82.9 1.5

Oil tanker 200000-+ 1 315981 215 2.70 0.0 0.0

20


Type Size Sample

size

Mean dwt Mean

days

at sea

Mean at

sea speed


allocative

utilization

Median

payload

utilization

Mean

transport

work

per ship


quartile

Upper

quartile

(%) (%) billion

t.nm

Bulk carrier 10000-34999 13 33355 74 13.24 14.22 12.93 16.64 79.5 77.6 0.5

Bulk carrier 35000-59999 4 54998 135 13.33 12.80 11.14 15.62 57.8 81.1 1.1

Bulk carrier 100000-199999 15 176128 105 12.81 6.63 6.17 7.43 48.4 95.4 2.5

Bulk carrier 200000-+ 1 205097 230 13.05 6.63 6.63 6.63 49.8 86.4 6.4

Chemical tanker 10000-19999 2 14583 160 10.07 50.07 46.16 53.98 79.0 51.2 0.2

Container 1000-1999 2 15016 150 15.00 56.99 55.86 58.12 100.0 70.0 0.6

Container 2000-2999 7 33582 249 16.97 29.44 26.46 32.64 100.0 70.0 2.4

LPG

0-49999 11 9134 120 13.28 145.47 98.92 160.74 47.0 41.5 0.1

LPG

50000-199999 1 64220 3 11.74 0.0 NaN

Oil tanker 120000-199999 11 155606 141 11.68 10.27 9.24 13.10 41.5 84.3 2.1

Oil tanker 200000-+ 1 298969 91 10.02 413.48 413.48 413.48 1.2 39.7 0.0

21


Type Size Sample

size

Mean dwt Mean

days

at sea

Mean at

sea speed


allocative

utilization

Median

payload

utilization

Mean

transport

work

per ship


quartile

Upper

quartile

(%) (%) billion

t.nm

Bulk carrier 10000-34999 15 33369 119 12.38 12.33 10.80 14.86 79.5 72.5 0.7

Bulk carrier 35000-59999 4 54998 130 13.01 12.32 11.41 14.06 75.1 69.6 1.1


199999

19 176692 112 12.51 6.44 5.84 7.46 47.3 93.9 2.1

Bulk carrier 200000-+ 3 205143 122 11.66 5.74 4.51 6.46 51.8 97.7 3.4

Chemical tanker 10000-19999 2 14583 120 8.95 66.83 33.68 99.99 70.2 58.5 0.2

Container 1000-1999 3 16702 128 15.19 46.21 32.64 49.67 100.0 70.0 0.5

Container 2000-29999 5 33641 231 15.60 23.10 21.37 31.14 100.0 70.0 2.0

LPG

0-49999 3 12044 1 12.98 61.66 61.66 61.66 0.0 47.0 0.0

LPG

50000-199999 1 54155 3 15.67 0.0 NaN

Oil tanker 120000-

199999

7 155693 141 11.84 8.22 7.72 14.46 41.9 83.4 2.0

22


Type Size Sample

size

Mean dwt Mean

days at

sea

Mean at

sea speed


allocative

utilization

Median

payload

utilization

Mean

transport

work

per ship


quartile

Upper

quartile

(%) (%) billion

t.nm

Bulk

carrier

10000-34999 15 33369 66 11.59 13.15 11.59 16.06 72.6 81.1 0.4

Bulk

carrier

35000-59999 4 54998 74 11.78 8.97 8.83 9.75 80.3 76.2 0.7

Bulk

carrier

100000-

199999

16 176248 82 11.95 6.62 5.86 7.23 42.9 93.3 1.6

Bulk

carrier

200000-+ 2 205167 152 11.06 5.03 4.84 5.21 46.4 97.8 3.8

Container 1000-1999 2 15016 86 14.07 51.02 42.35 59.69 100.0 70.0 0.3

Container 20000-2999 7 33582 139 15.76 25.94 24.15 27.66 100.0 70.0 1.2

Oil

tanker

120000-

199999

9 155562 114 10.29 11.53 8.48 12.77 43.1 77.2 1.1

23


Type Size Sample

size

Mean dwt Mean

days

at sea

Mean at

sea speed


allocative

utilization

Median

payload

utilization

Mean

transport

work

per ship


quartile

Upper

quartile

(%) (%) billion

t.nm

Oil tanker 120000-199999 10 154063 9 11 6.27 5.43 35.66 0 89.97 0.10

Oil tanker 200000-+ 3 307339 6 10 0 0.00

24

5. Comparison studies analysis

To prepare for the imminent EU MRV legislation, a number of studies have been

undertaken to measure operational energy efficiency using ship owner or vessel tracking

data. In this section, we compare our results to the following studies that had comparable

ship types8:

Marin, 2014: Towards a realistic CO2-MRV model for merchant shipping. MRV

study performed for the Dutch merchant fleet

Intertanko, 2013: Report from the ISTEC/Environmental committee joint working

group on MRV (JWG/MRV).

Kristensen, 2013: Experience with Energy Efficiency Operational Indicators

(EEOI) Seen in the Light of MRV. Danish Shipowners’ Association.

MEPC 68/INF.29. (2015): Empirical comparative analysis of energy efficiency

indicators for ships.

MEPC 68/INF.24. (2015): The Existing Shipping Fleet’s CO2 Efficiency.

Figure 6 shows a comparison of fuel consumption/nautical miles for a container ship as

reported in Marin (2014). In 2013, this indicator decreased to approximately 40% of its

2012 value (decreasing even further into 2014). In addition, the standard deviation as a

percentage of the mean is plotted. The large standard deviations indicate that the mean

values have little statistical meaning (Marin, 2014) and high variability on a day-by-day

basis. We compare the results of the Marin study of a single containership to our data

from this study. We plot a single container (bottom-left Figure 6) and all containerships

in the data sample at top of Figure 6. The data from this study for both the single

containership and the average for all containerships show very little change in fuel

consumption per unit distance over the period (2010 through to mid 2013), which

contrasts with the Marin study’s data’s steady decrease. The explanation is likely to be

that the ship which the Marin study is describing experienced significant changes to its

technology or operation (e.g. slow steaming) between 2012 and late 2014, whereas this

study’s containerships did not. The standard deviation for the individual ships studied

(both in the Marin study and the sample used from this study) is less than the standard

deviation of the combined containership fleet’s data, which implies that the variability of

one ship’s operation is greater than the variability across the fleet of ships in this study.

8 In some studies, the EEOI was not calculated for the ship types in our study. In these cases, we compared

our data with the metric used in the published study.

25

In combination, these comparisons indicate that the global fleet of a given ship type/size

was not modified in the same way over the period (there is some variation from one

owner/operator to another), as well as showcasing the efficacy of the fuel/dist indicator

for detecting basic trends in performance.

Figure 6 Comparison with MARIN study. Top: All containership data in the RBSA

sample; bottom left a containership in the sample; bottom right Marin study fig.7 data

from a container

26

Figure 7 compares the data in our study for oil tankers to data provided by Intertanko

members (Intertanko, 2013). While the EEOI decreases as DWT increases, CO2/distance

increases as DWT increases.

Figure 7 Comparison with INTERTANKO study. Top EEOI and CO2/distance from

INTERTANKO presentation JWG/MRV at Hellenic Mediterranean Panel 2013; bottom

EEOI and CO2/distance from our data sample for all tankers.

Generally, this study’s fleet contains a number of small tankers (below 32,000 DWT) for

which there are no equivalents in the INTERTANKO sample. There is only one size class

27

(150,000 DWT) for which the data can be reasonably compared. For that size, the

INTERTANKO data is on average lower in value and tightly clustered with EEOI’s

ranging from approx. 4-7 gCO2/t.nm. This contrasts with this study’s data which covers a

range of 6-15 gCO2/t,nm, and includes a number of outliers of significantly higher

values. The same is true of the CO2/total distance data, the INTERTANKO study fleet

has on average lower and more tightly clustered values than this study’s data. This

implies that overall, the fleet in the INTERTANKO study were on average either

technically more efficient, operationally more efficient, or both technically and

operationally more efficient. However, the presence of the outliers (particularly the high

outliers) in this study’s data suggests that another possible explanation is that the data

used in the calculations for this fleet’s EEOI and CO2/total distance contained some

unreliable or spurious values, which is consistent with comments made in Section 1 about

data quality.

We compare our data on bulk carriers to two studies. Figure 8 (top) plots the relationship

between fuel consumption and DWT as reported in MEPC 68/INF.29 (2015), a study

which compares the indicators that are currently under discussion at MEPC, namely the

Annual Efficiency Ratio, the Individual Ship Performance Indicator, the Fuel Oil

Reduction Strategy and the unnamed United States proposal. A logarithmic curve is fitted

to the data, providing a low R2 of 0.0548 and there is a considerable spread in the

observations that cannot be explained by DWT. We compare this representation to our

data for the bulk carriers in our sample (bottom plot). Although there is still considerable

variation that cannot be explained by DWT, both curves (logarithmic and power) have a

higher R2

(0.29) than the MEPC 68/INF.29. (2015) study, which cannot be explained by

the different curve fit. There is also a lower standard deviation in the comparison study

for ships with lower DWT than in our data sample, which shows a fairly consistent

standard deviation across different ship sizes of bulk carriers.

28

Figure 8 Comparison with MEPC 68/INF.29. (2015) study.

29

Another study by Kristensen (2013) provides curves of EEOI derived from model

calculations and EEOI data reported by ship owners (Figure 9 top). The ship owner

reported data is calculated in two ways, some data (the red dots) are for annual aggregates

of EEOI and the yellow dots are for a single month’s data. Both correspond to the year

2013. The ship owner data is reported as the average EEOI for a number of ships in a

certain size range, so there is no indication of the variability in the EEOI’s between

similar ships. The data from Kristensen (2013) can be compared with a presentation of

the data from this study (all years) for bulk carriers (Figure 9 bottom). Across the range

of sizes this study’s data shows, average EEOI’s higher than the data presented in

Kristensen (2013). For example, for ships around 50,000 DWT, the point estimates from

the Kristensen (2013) study show values of approximately 4-10 gCO2/t.nm, whereas this

study’s data provides values in the lower quartile of greater than 8 gCO2/t.nm. However,

inspecting the tabular results (Table 9) for the year 2013 shows that considering this year

in isolation, this study’s data has median values much closer to the average values in

Kristensen (2013) (for example median capsize EEOIs are 5-6.6 gCO2/t.nm, median

EEOI for 55,000 DWT fleet is 9gCO2/t.nm). Even on 2013 data only, this study’s data

still appear slightly higher than the Kristensen (2013) data, but now more credibly to do

with differences between operation specifics and specialization rather than a fundamental

difference between the fleets.

This comparison therefore shows the importance of drawing comparisons on a year-by-

year basis, particularly during the period for which this study’s data has been collected

(2007-2013). In particular, this is because of the increasing prevalence of slow-steaming

(seen clearly in the tabular data) that contributed to important increases in efficiency over

the period.

30

Figure 9 Comparison with Kristensen, 2013 (top) study to RBSA data (bottom)

A study using vessel tracking (AIS) data to estimate the EEOI is MEPC 68/INF.24.

(2015). This study was a follow-on to the IMO Third GHG Study 2014, and deployed

AIS derived estimations of fuel consumption in combination with AIS derived estimates

of cargo mass and transport work to calculate EEOI. The cargo mass is obtained from the

31

AIS reported instantaneous draught and a series of algorithms filters the results to remove

spurious data. The results present the statistics of transport work and EEOI broken down

by ship type and size category (defined by dwt size). The filtering to remove spurious

data leaves approximately 10-20% of the fleet that is assumed to be a representative

sample of the global fleet (the representativeness is tested as part of the study).

MEPC 68/INF.24 is used as a source for point estimates of the median EEOIs of all ship

types and sizes as reported in Table 11 for the same ship types and sizes in this study.

There is no reason that the sample of RBSA ships that were included in this study should

be representative of the overall fleet median, but similarity between the two numbers in a

given year provides some indicative reassurance on the quality of the data.

Table 11 Comparison of EEOI tabular results with MEPC 68/INF.24. (2015)

MEPC 68/INF.24. (2015). This study

Type Size 2010 2011 2012 2010 2011 2012

Bulk carrier 10000-

34999

13.4 14.4 15.4 14.36 14.22 12.33

Bulk carrier 35000-

59999

10.6 11.5 11.7 13.36 12.80 12.32


199999

12 6.34 5.83 7.65 6.63 6.44

Bulk carrier 200000-+ 8.85 5.41 5.13 7.12 6.63 5.74

Chemical

tanker

10000-

19999

22.1 23.2 23.7 - 50.07 66.83

Container 1000-1999 28.1 31 31.6 - 55.86 50.82

Container 2000-2999 21.1 24.6 24.7 30.22 30.52 28.11

LPG 0-49999 24.7 27.9 30.4 146.57 145.47 61.66

LPG 50000-

199999

13.9 15.3 16.3 1428.24 - -

Oil tanker 120000-

199999

14.3 9.12 10.8 11.99 10.27 8.22

Oil tanker 200000-+ 4.89 6.47 6.57 - 413.48 -

32

There is generally reasonable agreement in the magnitudes between MEPC 68/IINF.24

and this study with some notable exceptions:

In INF. 24, chemical tankers have lower EEOIs than those estimated in this study.

This could be attributable either to the quality of the data used in this study, or the

quality of data in INF. 24. Chemical tankers present a particular challenge for AIS

derived estimates of EEOI since there is often fuel consumed for cargo heating

and operations which AIS data can only estimate. They are also ships that operate

often part-loaded with cargo which makes them more challenging with respect to

the accuracy of AIS derived cargo mass. Inspecting the data further, INF 24

estimated utilization (the combination of payload and allocative utilization) is

nearly twice (1.75 times, in 2011 and similar in 2012) the magnitude of the

utilization estimate for the ships in this study. This goes some way to explaining

the difference observed, but even taking this into account the EEOI’s calculated

for this study are approximately 20% higher than the INF. 24 values if like for

like utilization had been measured. This implies that there is also a difference in

the fuel consumption – with this study’s ships reporting higher fuel consumption

than that estimated in the Third IMO GHG Study 2014. INF. 24 estimates that the

average speed and days at sea for this size category of chemical tankers are both

higher than the speed and days at sea for this study’s ships. Therefore the probable

explanation is a difference in the auxiliary fuel consumption. Without further

analysis it is not possible to say whether this is because the ships in this study

have some unique features and operation that require higher auxiliary fuel

consumption, whether the data quality is the source of the discrepancy, or whether

deficiencies in the method and assumptions in INF. 24 are the explanation.

Container ships appear to have lower EEOIs in INF. 24 than in this study.

However, in this study as no cargo data was available a single size-generic

assumption was applied for payload and allocative utilization. INF. 24 observes

that smaller container ships have higher utilization than larger container ships

which if applied to this study’s data would have the effect of lowering the EEOI’s

calculated for the two size categories for which time-series data was available.

This may resolve the observed discrepancy.

Liquid gas carriers have very large discrepancies between this study’s estimated

values and those from INF.24. The likely explanation is the cargo data used in this

study for this ship type - there were many inconsistencies and null values reported

in the raw data describing the cargo mass for this ship type, and no means to

estimate or validate or correct. The consequence of a shortage of cargo data is

unreliability in the calculation of transport work and a likely under-calculation of

33

transport work and this would result in an overestimation of EEOI, which is

consistent with the tabular results in Table 11.

There is reasonable agreement for the Suezmax tanker size class (120000-199999

DWT), however poor agreement for the VLCC tanker size class (200000 +

DWT). The explanation is likely to be the quality of the cargo data available in

this study for this size category of tanker.

Other than these specific larger discrepancies, there are a few discrepancies in trends over

the time period. For example some ship type and size categories in this study show

improving EEOIs with time when the INF. 24 results imply deteriorating EEOIs over the

same time period. This could be the result of a comparison between a single operator,

instilling operation efficiency optimization practices in their specific fleet, and the

median performance of the wider fleet in which that operator was based.

34

6. Decomposition analysis of EEOI

As discussed in Section 2, the EEOI can be decomposed into logistic and technical

factors. Logistic factors include the ship’s average payload utilization, allocative

utilization, and a factor reflecting operating speed and draught. The variables influence

both the numerator and denominator of the EEOI equation in different ways, making

interpretation of the index not completely straightforward.

A ship’s payload utilization affects both fuel consumption and transport work. A ship that

is carrying cargo requires more energy for propulsion at a given speed, compared to a

ship that is carrying less cargo or ballasting empty (fuel consumption). The cargo carried

over a time period also determines the tonnes component of transport work over the same

time period. Allocative utilization, or days in laden divided by the total sailing days,

affects the transport work or revenue miles through the distance component.

The operational speed affects the fuel consumption, and the transport work achieved

within a given period of time. It is a common error to assume that a ship that slow-steams

will always obtain a higher operational efficiency compared to an equivalent ship sailing

at a faster speed. Any differences in payload and allocative utilization also need to be

taken into account. In Figures 10-18, we examine these relationships for bulk carriers to

understand the main drivers of EEOI. Note that the allocative utilization shown in the

figures is the ratio of distance laden to total sailing distance. This metric will be the same

as the days laden to total sailing days (the allocative utilization) if the average operating

speed is the same as the average laden speed, and approximate if the two speeds are

similar. The laden distance to total sailing distance ratio is used in order to be consistent

with other studies.

We first describe the EEOI in terms of the logistics factors by ship type and size,

alongside the same information with the estimated EETI (see Appendix A.1). Viewing

the data by size class allows for some control over the influence of technical efficiency

that increases with size. We then compare the EEOI to some alternative indicators

proposed in the IMO.

The decomposition analysis results are presented as time-series for each ship type. In

some instances, the composition of the fleet of ships (within a given type and size)

changes over time (ships are scrapped or new ships are added). These changes to the fleet

can influence the observed trends and so the results included in the plots shown in

Section 6 are limited to a subset of the ships that are consistently operating for the

duration of the time-series.

For the comparison against some alternative indicators, the following are considered: the

Annual Efficiency Ratio or the carbon emissions divided by the product of deadweight

and distance sailed (proposed by Japan and denoted as AER), carbon emissions divided

35

by total hours of operation (proposed by the US but with joules of energy rather than

carbon emissions and denoted as EEUSI), and fuel consumption over distance (proposed

in MEPC-66/4/6, 2013).

6.1 Decomposition analysis for bulk carriers

Figures 10-15 show the EEOI, EETI and logistics factors for three size classes of bulk

carriers for which there was a sufficient sample size of ships per year to justify a time-

series and trend analysis.

For bulk carriers in size 2 (DWT range 10,000-34,999), Figure 10, the EEOI decreases

between 2008 and 2009 and then slightly increases between 2009 and 2010. The logistics

factors can be used to explain this trend, as the decrease in EEOI can be explained by a

lower speed, which counteracts the decrease in payload utilization and flat allocative

utilization between 2008 and 2009. An opposite trend occurred between 2009 and 2010

with payload utilization improving and allocative utilization deteriorating, while speed

continues to decrease. Therefore the allocative utilization overrides the positive impacts

of changes in payload utilization and speed on efficiency.

For the same size category, Figure 11 presents the logistics factors alongside the

estimated EETI. The resulting EETI trend is a gradual deterioration over time –

increasing by an average of approximately 10% over a 2-year period.

36

Figure 10 EEOI and logistics factors for bulk carriers size range 10,000-34,999 dwt; size

sample: 13 ships.

37

Figure 11 EETI and logistics factors for bulk carriers size range 10,000-34,999 dwt;

sample size: 13 ships.

Figure 12 and Figure 13 present results for bulk carriers in the size class 3 (DWT range

35,000-59,999). Over the four years for which consistent data is available, the EEOI

consistently decreases (energy efficiency improves). The logistic factors contributing to

this trend include significant increases in allocative utilization and slower speed. These

factors offset the decreasing trend in payload utilization during 2008-2010. However the

increase in payload utilization, in combination with increasing allocative utilization and

reducing speed, can clearly be seen as contributing to a more precipitous decline in EEOI

during 2010-2011. On average, over the period studied, the EETI increases (again by

approximately 10%), with a slight decrease between 2010 and 2011.

38

Figure 2 EEOI and logistics factors for bulk carriers size range 35,000-59,999 dwt; size

sample: 4 ships.

39

Figure 13 EETI and logistics factors for bulk carriers size range 35,000-59,999 dwt; size

sample: 4 ships.

As shown in Figure 14 and Figure 15, for the larger size class 5 (100,000-199,999), the

EEOI reduces over the period 2008-2010, before increasing in 2011. Again, this can be

explained by the interaction of the variations in allocative utilization, along with a

declining payload utilization and decreasing speed. The EETI over the same period

steadily increases with time, this time by approximately 15% over the four-year period of

the study.

40

Figure 14 EEOI and logistics factors for bulk carriers size range 100,000-199,999 dwt;

size sample: 8 ships.

41

Figure 15 EETI and logistics factors for bulk carriers size range 100,000-199,999 dwt;


6.1.2 Comparison to alternative indicators

Figures 16-26 present the calculation of energy efficiency indices for three of the ship

size categories of bulk carrier (size 2, 3 and 5), for which a year-on-year consistent

sample of ships could be identified. The results show that whilst similar trends are

apparent in the AER, fuel consumption/distance and EEUSI, these are not always

consistent with the trend in EEOI. This appears to be particularly the case for size 2 and

size 5 which see reductions in operational efficiency between 2009 and 2010, and

between 2010 and 2011 respectively, whilst the alternative indices were showing

continued improvements.

This difference between the EEOI trend and the trends of the alternative indices can be

explained by the fact that the alternative indices have no means to capture the influences

of varying payload or allocative utilization – logistics factors that in Section 6.1 were

shown to produce important variations in the EEOI.

42

Figure 16 EEOI and alternative indicators for bulk carriers size range 10,000-34,999

dwt; size sample: 13 ships.

43

Figure 17 EEOI and alternative indicators for bulk carriers size 35,000-59,999 dwt; size

sample: 4 ships.

44

Figure 18 EEOI and alternative indicators for bulk carriers size range 100,000-199,999

dwt; size sample: 8 ships.

6.2 Decomposition analysis for container ships

Figure 19 presents results for the container ships in size category 3, a small sample of 5

ships. Information on the ships’ design/reference speeds was not available and therefore

EETI calculations were not possible. Furthermore, there was no data describing the

specifics of the ship’s cargo, and so the allocative utilization was set to 1 and the payload

utilization set to 0.7 as default assumptions for all ships.

Over the four year time period for which a consistent time-series was possible, the EEOI

gradually improved, reducing by approximately 15%. This is driven by a reduction in

operating speed – in this instance with assumptions for cargo, no other logistics factors

could possibly induce variability in EEOI.

Figure 20 presents results comparing the different energy efficiency indices. As there is

no variation in the cargo over time or between ships, it is not surprising that unlike the

case for the bulk carrier fleet, all four indices display the same trend over time. This at

least demonstrates that each of the indices has a similar ability to detect variations in

45

operational efficiency due to changes in operating speed. This is because the numerator,

fuel consumption, is the same across all indices, and is affected by speed.

Figure 19 EEOI and logistics factors for containers size range 2000-2999 TEU size

sample: 5 ships.

46

Figure 20 EEOI and alternative indicators for containers, size range 2000-2999 TEU;


6.3 Decomposition analysis for Liquefied Petroleum Gas (LPG) carriers

Figure 21 and Figure 22 present trends in EEOI, EETI and logistics factors for a subset of

the LPG carrier’s fleet (size 1, 0-49,999 CBM capacity) over the period 2008-2010.

There is a wide variability in the EEOI of the fleet (between ships), which appears to be

largely driven by the variability in the fleet’s payload utilization – some ships carry, on

average, half as much cargo as other similar ships in the same fleet.

Consistent with other ship types and sizes, the fleet shows a steady improvement in EEOI

over the period of the study, driven both by a lowering of operating speed, and an

increase in allocative utilization (the median payload utilization stays approximately

constant with time).

The trend in EETI is for a steady increase, again consistent with the other ship types and

sizes – in this instance a deterioration of approximately 10% over the period 2008-2010.

47

Figure 21 EEOI and logistics factors for LPG carriers, size range 0-49999 cbm; size

sample: 6 ships.

48

Figure 22 EETI and logistics factors for LPG carriers size range 0-49999 cbm; size

sample: 6 ships.

Figure 22 shows that in this instance, the different energy efficiency indices give slightly

different trends. EEOI improves over the period of the study (by a small amount),

whereas the other three indices either deteriorate or stay approximately constant. As with

other ship types, this can be explained as the result of the EEOI capturing variations in

logistics factors that are not considered in the other three indices.

49

Figure 23 EEOI and alternative indicators for LPG carriers size range 0-49999 cbm;


6.4 Decomposition analysis for oil tanker

Figure 24 and Figure 25 present results for the fleet of size 7 (DWT range of 120,000 –

199,999) tankers, a fleet of three ships with consistent data year on year during 2010 –

2012. Reducing average speeds, increasing allocative utilization and reducing payload

utilization all trade off to create an initial improvement in EEOI (2010-2011), followed

by a deterioration (2011-2012) and a net deterioration over the period 2010-2012.

Over the same period, the EETI deteriorates initially, and then improves so that in 2012 it

is similar in magnitude to its value in 2010.

For the energy efficiency metrics (Figure 26), in this instance, both fuel

consumption/distance and EEJI provide similar trends, whereas the EEUSI shows a

consistent downward trend. Only the EEOI shows a net deterioration during 2010 to

2012, which is driven by the significant reduction in payload utilization between 2011

and 2012.

50

Figure 24 EEOI and logistics factors for oil tankers size range 120000-199999 dwt; size

sample: 3 ships.

51

Figure 25 EETI and logistics factors for oil tankers size range 120000-199999 dwt; size

sample: 3 ships.

52

Figure 26 EEOI and alternative indicators for oil tankers size range 120000-199999 dwt;


6.5 Summary

In summary, taking the results for all different ship types and sizes:

There is little evidence of correlation between EEOI and any one logistics factor,

for all of the types and sizes considered, variations in EEOI can be explained only

by considering contributions from a combination of the logistics factors.

Consistently across all ship types and size categories, speed decreased during the

period of this study, trends for both allocative and payload utilization differ and in

many cases showed deterioration in utilization which counteracts operational

efficiency improvements obtained by slow-steaming.

On average, the trend is for moderate decreases (improvements) in EEOI during

the period of the study.

In general, the variability in EETI in any given year and ship size, appears smaller

than the variability in EEOI. This appears particularly true for the larger ships

(e.g. Capesize), which is to be expected since these ships are more likely to be

technically homogenous to begin with. This is to be expected as the drivers of

53

EEOI should have greater variability than the drivers of EETI (hull, propulsion

and machinery condition changes, and metocean variability).

The estimated trend in EETI appears consistent with expectations that

performance is likely to deteriorate over time (e.g. coating and fouling

deterioration, propeller damage, engine wear), and implies that EEOI

improvements are being obtained over the period of the study (reducing EEOI

with time) in spite of this underlying technical efficiency deterioration – mainly

due to extensive implementation of slow steaming.

However, as there are some other trends that are well correlated with the EETI

(e.g. reducing speed, increasing EETI) it is possible that the trend in EETI is the

result of an inaccuracy in its estimation (e.g. due to an inadequacy in the power 3

used in the speed factor derivation). Further work would be needed with data

controlling for the trend in speed in order to understand this better.

For each of the different ship types and sizes, the alternative operational energy

efficiency indices show varying levels of trend agreement/disagreement.

Significantly, it is often the case that the three alternative indicators produce

different trends to EEOI. This shows that a) no alternative energy efficiency

metric is a reliable proxy to EEOI and b) the choice of energy efficiency metric is

a function of what information is believed to be of greatest importance. Of the

indicators considered, the EEOI is the only indicator that represents the

carbon intensity of the actual transport work done (when measured in

t.nm’s), all other indicators approximate transport work done in some way.

7 Factors influencing the EEOI

7.1 Interviews with companies on technical and operational measures

The decomposition analysis from Section 6 highlighted that the EEOI is influenced by

technical and logistics factors. Because we were unable to measure the technical

efficiency of the ships in the data sample in their real operating conditions nor obtain data

on any retrofits that might have occurred, we conducted interviews with some of the

owners for the study to see if any technical interventions were undertaken during the

sample period (2008-2012).

For the liquefied gas products (LPG), no intervention measures were undertaken during

this period. However, a consultant working on the technical performance of the LPG fleet

commented that investment in technical measures were implemented in 2012, therefore

post 2012 it would be likely that a step change occurred in the technical efficiency of

their LPG fleet which would not be seen in the data we analyzed.

For the bulk carrier ships in the sample, the ships have Propeller Boss Cap Fins, and two

have Mewis ducts. The Supramax ships have trim optimization.

54

For container ships, the technical manager reported that they are not investing in energy

efficiency improvements because the ships are time-chartered and they “don’t receive a

higher time charter rate.” This split incentive arises in the time charter market because the

fuel costs are borne by charterers (in addition to the daily charter rate) and capital and

maintenance costs are borne by the ship owner or operator. This represents a type of

principal-agent problem that also arises in the building sector between a tenant and

landlord (see Blumstein et al. (1980), Scott (1997), Maruejols & Young (2011), Bird &

Hernandez (2012)).

The degree of the split incentive in each market segment depends on the size of the time

charter market, the length of the contracts and whether charterers are willing to reward

owners for their investments in energy efficiency or clean technologies. Estimates from

an analysis of fixtures in 2011 (Rehmatulla, 2014) shows the size of the time charter

market varies in each sector. In the tanker sector, the majority (around 90%) of ships are

traded on the voyage charter (spot) market, while in the dry bulk sector time charter

contracts could account for as much as 60% of the ships, suggesting that this type of

barrier is a larger problem in the dry bulk sector.

The extent to which the fuel cost savings are passed back to the owner-operator through

higher charter rates will create direct incentives for ship owners and operators to

implement wind technology. Agnolucci, Smith & Rehmatulla (2014) show that on

average only 40% of the financial savings delivered by energy efficiency accrue to the

ship owner for the period 2008–2012 in the dry bulk Panamax sector. The incomplete

pass-through of savings through higher rates has also been suggested in Smith et al.

(2013a), Riise & Rødde (2014) and Parker & Prakash (Accepted) and has been referred

anecdotally in Faber, Behrends & Nelissen (2011), Maddox Consulting (2012) and

Lloyds List (2013).

In terms of operational measures, the companies interviewed are using voyage

optimization in varying degrees. For the liquefied petroleum gas carriers in the sample,

voyage optimization is used to optimize for current and weather due to the performance

fuel claims. For example, mid-size LPG travel on long-haul routes (i.e., Arabian Gulf to

India) so they can benefit more from interventions and are taking advantage of favorable

currents. The company is starting to monitor and train captains to create awareness that

fuel consumption can vary due to the crew’s handling of the ship and to collect consistent

noon reports to track the fuel consumption on ships.

For the bulk carriers in the sample, super slow steaming was implemented from 2013,

with ships sailing at economical speeds (45% of MCR) from 2008 and this was verified

by the data presented in Section 6. Voyage optimization is only used if the masters ask

for it, but “it is not always accurate, sending some ships on bad routes (to rocks,

islands).”

55

The companies were receptive to understanding what data might be useful to collect in

the future for the EEOI calculations so that they can optimize the collection process.

7.2 Time series of freight rates, bunker prices, and speed

The financial crisis of late 2007 lead to a recession in the United States in December

2007 and spread globally, impacting freight rates at the end of 2008. Figure 27 depicts the

freight rates and bunker prices as moving in tandem until late 2008 as a result of the

global recession, but the series diverge after 2009 when bunker prices continued to rise

while freight rates were still dampened by the effects of an oversupplied market.

Figure 27: Dry bulk time charter rates and bunker prices (2008-2013). Source: Clarkson

Research.

56

Figure 28: Average sailing speed of bulk carrier fleet 2007-2012. Source: Smith et al.,

2013

Speed is related to market conditions, with vessel speeds increasing as market conditions

improve and the value of time becomes more important relative to cost savings. This

relationship can be seen from vessel tracking data (Smith et al., 2013) in Figure 28, which

shows the average speeds for bulk carriers by size class. Average speeds peaked in 2008,

with the exception of the Capesize segment which peaked in 2009. This largely reflects

the trend in mean speeds of the bulk carriers in the RBSA sample, which dropped from

14.6 to 14.0 between 2008 and 2009, remaining flat in 2010, and then steadily decreasing

from 14.1 to 11.7 knots between 2010 to 2013.

Poor market conditions also mean that ships have a lower probability of obtaining cargo

fixtures per time period, and may have to sail further in search of cargo. During a

depressed market, ship owners do not have bargaining power and have to accept lower

payloads in order to avoid ballasting empty to another destination. This can be seen in the

overall bulk carrier data, which saw its median payload utilization drop each year starting

in 2009, from 97% to 84%. Allocative utilization, which is size specific due to the fact

that different ship sizes are optimized for specific bulk trades and thus will differ

depending on the distance between load and discharge areas, varied across time and size.

There was a decreasing or flat trend for size classes 2 and 5 and an improvement for size

class 3.

57

An improvement in allocative utilization may signify that fuel costs dominate

owner/operators’ decisions about where to allocate their ships due to the low or negligible

profit margin earned. However, it is also likely that owners had to ballast further in search

of employment, thus leading to a decrease in the allocative utilization. We also did not

consider changes in trade patterns during this period, which could also alter allocative

utilization.

Figure 29: Tanker rates and oil prices (2009-2014)

The tanker sector has also experienced a depressed market during the study’s period

(2008-2014) due to oversupply, although conditions improved slightly in 2014 as can be

seen in Figure 29 for oil tankers. For oil tankers in size class 7, the sample shows a

decline in payload utilization between 2011 and 2013 and then an improvement in 2014,

while the allocative utilization dropped after 2010, remaining moderately flat through to

2013 (there was no data available in 2014). Median speeds fluctuated around 11 knots

between 2010 and 2013, increasing to 12 knots in 2014.

For the two chemical tankers analyzed in the dataset, the median speed decreased

between 2011 and 2012 (10.1 to 9.0 knots), while payload utilization improved slightly

and allocative utilization decreased.

For LPG, the general trend was a decrease in speeds from 14.8 in 2008 to 12.5 knots in

2012 and payload utilization decreasing during this time period. At the same time,

allocative utilization improved slightly.

58

Figure 30: Container rates (2009-2015). Source: Shanghai Shipping Exchange;

Economist.com

The economic picture for the container sector is mixed, with two peaks in rates in 2010

and 2012 as depicted in Figure 30 (there were no reliable price indices before the

financial crisis (Economist, 2015)). Because of the rising price of fuel and low rates

during the financial crisis, container operators made strides in improving the efficiency of

their fleets, replacing old vessels for bigger, more fuel-efficient ones to be more resilient

to poor market conditions (Economist, 2015).

For the container ships in the sample, the median speed decreased between 2010 to 2012

from 17.4 to 15.7 knots, increasing to 16 knots in 2013. As described in Section 6, data

was not available on payload utilization or allocative utilization.

7.3 Laden EEOI and contract type

59

Ship type IMO size Contract type

Observations Median laden EEOI

Bulk carrier all 0 180 3.21

Bulk carrier all 1 2408 6.77

Bulk carrier 2 0 46 7.20




Bulk carrier 4 0 2 -




Bulk carrier 6

0 17 2.10

Bulk carrier 6

1 71 2.98

LPG all 0 300 62.25

LPG all 1 3210 49.95

LPG 1 0 300 62.25

LPG 1 1 3170 49.77

LPG 2 0 0 -

LPG 2 1

40 54.91

Table 12: Median laden EEOI by contract type where 0=SPOT and 1=Time charter,

2008-2013

We analyze the laden EEOI by contract type for a subset of the data as it was only

available for bulk carriers and LPG carriers. Table 12 reports the median laden EEOI by

contract type. For both ship types, over 90% of contracts were time charter.

60

By ship size, classes 2, 5, and 6 have a lower median laden EEOI for spot market

contracts compared to their respective size classes in time charter, while class 3 has a

higher laden EEOI in the spot market compared to the time charter market. The classes

with a lower laden EEOI in the spot market can be partially explained by ships being

operated at slower speeds over the time period; the median speed for all bulk carriers was

11.6 knots for ships operated on spot compared to 13.1 knots on time charter. Payload

utilization was .04 percentage points higher for the spot market compared to time charter

market.

In the LPG sector, the opposite trend occurred, with the laden EEOI higher for tankers on

spot compared to time charter. This is attributed to a lower payload utilization in the spot

market.

7.4 Summary

This section provided details on the possible technical and economic drivers of the sub-

indices presented in Section 6. From interviews with the companies owning bulk, LPG,

and container ships, there were no major technical interventions during the period of the

study. However, the company owning bulk carriers was taking operational measures such

as economical (slow) steaming and voyage optimization, while the LPG fleet was

responding to fuel performance claims by better monitoring the ship’s fuel performance

and considering retrofits.

Given the bleak prospects for the shipping industry at the end of 2008 (the first year of

the study’s time period) and high fuel prices through to 2013, ships were slow steaming

for all ship types in the study until at least 2013. Payload utilization also followed a

similar declining trend, but there was no obvious pattern in allocative utilization. This

could be due to different strategies for owners, with some owners minimizing ballast

distance whereas other owners were forced to search for employment by ballasting

farther. Furthermore, we did not investigate whether there were changing trade patterns

which could also alter allocative utilization.

We found that there were significant differences in the laden EEOI by contract type for

the ship types we could investigate. For bulk carriers, the median laden EEOI for the

majority of size classes operated on spot contracts is lower than for ships operated on the

time charter market. This can partly be explained by the lower speed of bulk ships

operated in the spot market compared to time charter, and points to the fact that owners

which leased their ships on time charter obtained higher (less efficient) EEOI ratings than

owner-operated ships.

8 Conclusion

Technical and logistics factors are the key drivers of the Energy Efficiency Operational

Index (EEOI). This paper has mathematically decomposed the EEOI into sub-indices in

61

order to understand the drivers of the index. We find that there is a relationship between

technical efficiency (EEDI) and EEOI across different ship sizes, but there is a wide

dispersion of EEOI values within a ship size class. This can be explained by the variation

in logistics factors, with little evidence of correlation between EEOI and any one logistics

factor. For all of the types and sizes considered, variations in EEOI can be explained only

by considering contributions from a combination of the logistics factors.

We find that consistently across all ship types and size categories, speed decreased during

the period of this study, trends for both allocative and payload utilization differ and in

many cases showed deterioration in utilization which counteracts operational efficiency

improvements obtained by slow-steaming. We also found significant differences in the

laden EEOI when a ship was operated on the spot market compared to time charter for a

subset of the ships for which there was contract data, resulting in higher laden EEOI for

ships on time charter. This points to the importance of principal agent problems in the

shipping industry, in which the owner may not be in control of the operation and hence

efficiency of the ship for the majority of the operation of a ship.

We compared the EEOI and alternative indicators proposed at the IMO. For each of the

different ship types and sizes, the alternative operational energy efficiency indices show

varying levels of trend agreement or disagreement. Significantly, it is often the case that

the three alternative indicators produce different trends to EEOI. This shows that no

alternative energy efficiency metric is a reliable proxy to EEOI and the choice of energy

efficiency metric is a function of what information is believed to be of greatest

importance. Of the indicators considered, the EEOI is the only indicator that

represents the carbon intensity of the actual transport work done (when measured

in t.nm’s), all other indicators approximate transport work done in some way.

References

Agnolucci, P, Smith, T & Rehmatulla, N 2014, 'Energy efficiency and time charter rates:

Some evidence quantifying the extent of split incentive problem in the panamax drybulk

market', Transportation Research Part A: Policy and Practice, vol 66, pp. 173-184.

Blumstein, C, Krieg, B, Schipper, L & York, C 1980, 'Overcoming social and

institutional barriers to energy conservation', Energy, vol 5, pp. 355 – 371.

Bird, S & Hernandez, D 2012, 'Policy options for the split incentive: Increasing energy

efficiency for low-income renters', Energy Policy, vol 48, pp. 506-514.

Germanischer Lloyd SE. 2013. Guidelines for Determination of the Energy Efficiency

Design Index.

62

Hasselaar, T.W.F. Towards a realistic CO2-MRV model for merchant shipping. MRV

study performed for the Dutch merchant fleet; 2014, Nov. Report No. 28065-1-TM.

Intertanko, 2013. Report from the ISTEC / Environmental committee joint working group

on MRV (JWG/MRV). Hellenic Mediterranean Panel, Athens, October 10th 2013.

Kristensen, 2013. Experience with Energy Efficiency Operational Indicators (EEOI) Seen

in the Light of MRV. Danish Shipowners’ Association.

Lloyds List 2013, Who pays and who wins with fuel-saving retrofits?, viewed 07 January

2013, <http://www.lloydslist.com/ll/sector/ship-operations/article414528.ece>.

Maddox Consulting 2012, Analysis of market barriers to cost effective GHG emission

reductions in the maritime transport sector, viewed 12 December 2012,

<http://bookshop.europa.eu/en/analysis-of-market-barriers-to-cost-effective-ghg-

emission-reductions-in-the-maritime-transport-sector-

pbML3213436/;pgid=y8dIS7GUWMdSR0EAlMEUUsWb0000a_8LQxNb;sid=27eLfRn

Kx26Lbkvc5PAV2nvv3GSNrrsJp-c=?CatalogCategoryID=lPKep2Ix.

Maruejols, L & Young, D 2011, 'Split incentives and energy efficiency in Canadian

multi-family dwellings', Energy Policy, vol 39, pp. 3655-3668.

MEPC 68/INF.29. (2015). Empirical comparative analysis of energy efficiency indicators

for ships. In MEPC.

MEPC 68/INF.24. (2015). The Existing Shipping Fleet’s CO2 Efficiency. In MEPC.

Parker, S & Prakash, V Accepted, 'Risk or reward: Energy efficiency premiums in the

time charter market', ECONSHIP 2015, 24 - 27 June, Chios.

Scott, S 1997, 'Household energy efficiency in Ireland: A replication study of ownership

of energy saving items', Energy Economics, vol 19, pp. 187-208.

http://www.poten.com/wp-content/uploads/2015/04/Tanker_Opinion_20140829.pdf>.

Smith, T, O'Keefe, E, Aldous, L & Agnolucci, P 2013a, Assessment of shipping’s

efficiency using satellite AIS data, viewed 15 April 2013,

<http://www.theicct.org/sites/default/files/publications/UCL_ship_efficiency_forICCT_2

013.pdf>.

Rehmatulla, N 2014, 'Market failures and barriers affecting energy efficient operations in

shipping (unpublished PhD thesis)', UCL Energy Institute, University College London

(UCL), London.

http://www.lloydslist.com/ll/sector/ship-operations/article414528.ece

http://bookshop.europa.eu/en/analysis-of-market-barriers-to-cost-effective-ghg-emission-reductions-in-the-maritime-transport-sector-pbML3213436/;pgid=y8dIS7GUWMdSR0EAlMEUUsWb0000a_8LQxNb;sid=27eLfRnKx26Lbkvc5PAV2nvv3GSNrrsJp-c=?CatalogCategoryID=lPKep2Ix.




http://www.poten.com/wp-content/uploads/2015/04/Tanker_Opinion_20140829.pdf

http://www.theicct.org/sites/default/files/publications/UCL_ship_efficiency_forICCT_2013.pdf

http://www.theicct.org/sites/default/files/publications/UCL_ship_efficiency_forICCT_2013.pdf

63

Riise, S & Rødde, M 2014, 'The relationship between ship specifications and time charter

rates: Quality premiums through a generalized additive model (unpublished MSc thesis)',

Norwegian School of Economics, NHH, Bergen.

Rojon, I & Smith, T 2014, On the attitudes and opportunities of fuel consumption

monitoring and measurement within the shipping industry and the identification and

validation of energy efficency and performance intervention, viewed 24 February 2014,

<http://www.bartlett.ucl.ac.uk/energy/news/documents/Report-

On_the_attitudes_and_opportunities_for_fuel_consumption_monitoring.pdf>.

The Economist 2015, Low Rates on High Seas, viewed April 1, 2015,

http://www.economist.com/news/business/21646234-slump-shipping-rates-reflects-

chronic-optimism-shipowners-low-rates-high-seas

VITO, 2009. Development of a uniform CO2 indicator system for the Belgian merchant

fleet.

http://www.bartlett.ucl.ac.uk/energy/news/documents/Report-On_the_attitudes_and_opportunities_for_fuel_consumption_monitoring.pdf

http://www.bartlett.ucl.ac.uk/energy/news/documents/Report-On_the_attitudes_and_opportunities_for_fuel_consumption_monitoring.pdf

http://www.economist.com/news/business/21646234-slump-shipping-rates-reflects-chronic-optimism-shipowners-low-rates-high-seas

http://www.economist.com/news/business/21646234-slump-shipping-rates-reflects-chronic-optimism-shipowners-low-rates-high-seas

64

Appendix

Appendix A.1 Operational efficiency: the EEOI

The EEOI, or annual fuel consumption divided by transport work, can be considered as

the annual average efficiency of a ship in its real operating condition, taking into account

actual speeds, draughts, capacity utilization, distance travelled, and the effects of hull and

machinery deterioration and weather.

The metric used for quantifying the operational efficiency of shipping is the EEOI (IMO

MEPC.1/Circ.684, 2009):

(1)


𝐹𝑗𝑖

∑ 𝑚𝑖𝐿

𝑖 𝐷𝑖𝐿

where:

i = the voyage

j = the fuel type

𝐹𝑖𝑗= the amount of fuel consumed for the voyage i and fuel type j

𝐶𝑗𝐹= the carbon factor for fuel type j

𝑚𝑖𝐿= cargo mass of voyage i

𝐷𝑖𝐿= distance travelled in loaded voyage i

𝑣𝑖𝐿= average loaded speed of voyage i in nautical miles per hour

Each individual voyage i is summed over a time period t. Distance, 𝐷𝑖𝐿, equals the

product of 𝑑𝑖𝐿, the duration in days of loaded voyage i, 𝑣𝑖

𝐿 the average speed during the

loaded voyage (in knots) and 24. Substituting these terms in for 𝐷𝑖𝐿:


𝐹𝑗𝑖

∑ 𝑚𝑖𝐿

𝑖 𝑑𝑖𝐿𝑣𝑖

𝐿24

Efficiency is increasing as EEOI decreases. Equivalently, instead of expressing the EEOI

as summations of each voyage’s parameters, it can be expressed using the average

characteristics of a number of parameters during the course of a time period t (commonly

1 year) as:

(2)

65

𝐸𝐸𝑂𝐼 =𝐶𝐹(𝐹𝑑𝑡

𝐿 𝑑𝑡𝐿 + 𝐹𝑑𝑡

𝐵 𝑑𝑡𝐵 + 𝐹𝑑𝑡

𝑃 𝑑𝑡𝑃)

𝑚𝑡𝐿𝑑𝑡

𝐿𝑣ℎ𝑡𝐿 24

where:

𝐶𝐹= the average carbon factor9

𝐹𝑑𝑡𝐿 = the loaded average daily operating fuel consumption during period t

𝐹𝑑𝑡𝐵 = the ballast average daily operating fuel consumption during period t

𝐹𝑑𝑡𝑃 = the port average daily operating fuel consumption during period t

𝑑𝑡𝐿= the days a ship is sailing loaded during period t

𝑑𝑡𝐵= the days a ship is sailing ballast during period t

𝑑𝑡𝑃= the days a ship is in port during period t

𝑚𝑡𝐿 = the average cargo mass during period t

𝑣ℎ𝑡𝑙 = the average loaded speed per hour h in time period t

Equation (2) can be further decomposed into the EEOI at sea and the EEOI in port:

𝐸𝐸𝑂𝐼 =𝐶𝐹(𝐹𝑑𝑡

𝐿 𝑑𝑡𝐿 + 𝐹𝑑𝑡

𝐵 𝑑𝑡𝐵)



+𝐶𝐹(𝐹𝑑𝑡

𝑃 𝑑𝑡𝑃)



𝐸𝐸𝑂𝐼 = 𝐸𝐸𝑂𝐼𝑠𝑒𝑎 + 𝐸𝐸𝑂𝐼𝑝𝑜𝑟𝑡

2.2 Technical efficiency in real operating conditions

Analogous to the use of the EEOI to estimate the operational efficiency of a ship, the

technical efficiency of a ship or energy efficiency technical indicator (EETI) can be

defined as the energy efficiency (gCO2/tonne-nautical mile) of a ship in a reference

operating condition (speed and draught):

(3)

𝐸𝐸𝑇𝐼 =𝐶𝐹𝐹𝑑

𝑟𝑒𝑓

𝑣𝑟𝑒𝑓𝐷𝑊𝑇24

where:

𝐶𝐹 = the average carbon factor of the fuel used


= the daily fuel consumption at a reference speed and draught

𝑣𝑟𝑒𝑓= the reference speed10

𝐷𝑊𝑇 = the deadweight tonnage of a ship

9 We assume that 𝐶𝐹 is the same for the loaded, ballast and port days. 10 The reference speed may or may not equal the design speed.

66

The EETI is a ship’s estimated technical efficiency in real operating conditions at a

specific point in time, whereas the EEDI (gCO2/t-nm), is the ship’s design technical

efficiency at the start of its life and under specific EEDI assessment conditions.

Differences between a ship’s EEDI and EETI could arise due to fouling, modification of

technical specifications (such as retrofitting) or because the EEDI trial performance

cannot be recreated in real operating conditions.

In order to estimate the fuel consumption in a reference condition, a correction factor can

be applied to measurements of fuel consumption in a specific operating speed and

draught, 𝐹𝑑𝑜𝑝

.

(4)


=𝐹𝑑

𝑜𝑝

(∑ (𝑣ℎ𝑡

𝑜𝑝,𝑖

𝑣𝑟𝑒𝑓)

𝑛

(𝑇𝑜𝑝,𝑖

𝑇𝑟𝑒𝑓)2/3𝑝𝑖=1 ) 𝑝⁄

where:

𝑇𝑜𝑝,𝑖 = the operating draught at passage or passage segment i

𝑇𝑟𝑒𝑓 = the reference draught

𝑣ℎ𝑡𝑜𝑝,𝑖

= the average operating speed for a passage or passage segment i (nautical

miles/hour).

p = total number of passages or passage segments

Approximating fuel consumption in real operating conditions raises issues around: the

accuracy of the correction factor formula (and its applicability to a specific design) and

the allowance for the influence (or correction for) weather and metocean conditions

(wind, waves, currents) on fuel consumption.

The former issue (applicability of the correction factor) could be addressed by using ship

specific correction formulae derived from sea trials or model tests. These were not

available for this study, and so standard assumptions were used (applicability of the

Admiralty formula, and the use of n = 3, as justified to be generally applicable for most

tankers and bulk carriers in Smith et al. 2014).

The latter issue (effect of weather and metocean) cannot easily be resolved. Either the

data could be collected inclusive of any effects, or it could be filtered to contain only fuel

consumption measurements obtained during a day with weather and current effects

deemed to be sufficiently low as to be negligible. The inclusion of weather and current

effects is more ‘honest’, but route and seasonal effects will influence the EETI. The

67

route/season specificity will diminish if data is collected and averaged over a sufficiently

long period of time.

Both of these issues are linked to long-running debates in the operation and regulation of

ships and deserve greater attention than can be afforded here. For the purposes of this

study, we take the EETI to be inclusive of all weather and metocean effects encountered

by a ship in its day-to-day operation.

2.3 Decomposing EEOI into technical and logistics factors

The EEOI and the EETI, when estimated from measurements of a ship’s daily fuel

consumption and activity, are inextricably linked because they have overlaps in their

input data. The mathematical derivation of the two formulae can be used to show how

EEOI can be decomposed into a number of technical and logistics factors, one of which is

the ship’s EETI. This is useful as a means to break EEOI down into a series of drivers

that each influence the overall value. As the only ‘technical factor’, the EETI indicates

the technical efficiency contribution to the EEOI quantification, while three logistics

factors indicate the contribution of the specifics determining the ship’s commercial

operation (speed and utilization).

An annual estimate of EEOI normally includes emissions when a ship is in port and at

sea. Although the port emissions should be included in the total EEOI, for the purposes of

decomposition analysis it is important to separate the port emissions from sea emissions,

as they represent a source of variability in overall emissions which has little or no

connection to technical efficiency or the main logistics drivers of changes in operational

efficiency. Distinguishing between port-time emissions and sea emissions, also allows

port emissions to be effectively monitored.

As port emissions represent only a small fraction (e.g. as shown in Figure 5, less than

10%), a simplifying assumption can be made in order to reduce the complexity of the

computations:

𝐸𝐸𝑂𝐼𝑠𝑒𝑎 ~𝐸𝐸𝑂𝐼

This assumption is applied whilst combining (4) with (2) and inserting (3) (under the

assumption that n = 3), which enables the production of an expression that makes the link

between EEOI and EETI:

(6)

𝐸𝐸𝑂𝐼~

𝐶𝐹𝐹𝑑𝑟𝑒𝑓

((∑ (𝑣ℎ𝑡

𝑜𝑝,𝑖

𝑣𝑟𝑒𝑓)

𝑛

(𝑇𝑜𝑝,𝑖

𝑇𝑟𝑒𝑓)2/3𝑝𝑖=1 ) 𝑝⁄ ) (𝑑𝑡

𝐿 + 𝑑𝑡𝐵)



68

~

𝐶𝐹𝐹𝑑𝑟𝑒𝑓

((∑ (𝑣ℎ𝑡

𝑜𝑝,𝑖

𝑣𝑟𝑒𝑓)

𝑛

(𝑇𝑜𝑝,𝑖

𝑇𝑟𝑒𝑓)2/3𝑝

𝑖=1 ) 𝑝⁄ ) (𝑑𝑡𝐿 + 𝑑𝑡

𝐵)𝑣𝑟𝑒𝑓𝐷𝑊𝑇24

𝑣𝑟𝑒𝑓𝐷𝑊𝑇24𝑚𝑡𝐿𝑑𝑡


~𝐸𝐸𝑇𝐼

((∑ (𝑣ℎ𝑡

𝑜𝑝,𝑖

𝑣𝑟𝑒𝑓)

𝑛

(𝑇𝑜𝑝,𝑖


𝑣ℎ𝑡𝐿


𝐵)

𝑑𝑡𝐿

𝐷𝑊𝑇

𝑚𝑡𝐿

~𝐸𝐸𝑇𝐼

𝑚𝑡𝐿

𝐷𝑊𝑇

𝑑𝑡𝐿


𝐵)

𝑣ℎ𝑡𝐿

((∑ (𝑣


𝑣𝑟𝑒𝑓)

𝑛

(𝑇𝑜𝑝,𝑖

𝑇𝑟𝑒𝑓 )2/3𝑝𝑖=1


Rearranging for 𝐸𝐸𝑇𝐼:

(7)

𝐸𝐸𝑇𝐼~𝐸𝐸𝑂𝐼𝑚𝑡

𝐿

𝐷𝑊𝑇

𝑑𝑡𝐿


𝐵)

𝑣ℎ𝑡𝐿

((∑ (𝑣ℎ𝑡

𝑜𝑝,𝑖

𝑣𝑟𝑒𝑓)

𝑛

(𝑇𝑜𝑝,𝑖


Equations (6) and (7) show that it is possible to formulate EEOI or EETI in terms of one

another if a number of extra details about speed, cargo and the loaded/ballast voyages are

known. These are, as represented in the right hand side of (6) and (7):

𝑚𝑡𝐿

𝐷𝑊𝑇= the average payload utilization

𝑑𝑡

𝐿


𝐵)= the allocative utilization or ratio of laden days to total operating days

𝑣ℎ𝑡𝐿

((∑ (𝑣


𝑣𝑟𝑒𝑓)

𝑛

(𝑇𝑜𝑝,𝑖

𝑇𝑟𝑒𝑓)2/3𝑝𝑖=1


= the speed and draught factor

The speed and draught factor shows a moderate degree of complexity. That is because the

correction of speed and draught incur non-linear relationships (speed to the power 3, and

draught to the power 2/3). As a result of this non-linearity, simplifying to annualized

averages of operating speed and draught would create inaccuracies. The significance of

those inaccuracies will be a function of the variability in operating speed and draught

from one voyage to the next. The variability in operating speed is more important than

69

the variability in draught (as speed varies according to a higher power). For the purposes

of monitoring and reporting, it is possible to calculate the speed factor term on a voyage-

by-voyage basis that can be used to calculate a single annualized figure (e.g. there is no

need for voyage-level data to be reported).

In this study, draught information was not consistently available. Furthermore, it can be

seen that draught is of lower importance to the correction factor than speed. As a result

the draught correction is omitted, and only operating speed corrections (calculated for

each unique component sea passage) are included.

The average payload utilization is always less than 1, the allocative utilization is also

always less than one, and the speed factor could be greater than, equal to, or less than 1.

Although commonly, especially recently, it will be greater than 1 – demonstrating slow

steaming.

In operation, a ship with a higher EETI value can offset this lower efficiency

disadvantage by obtaining a higher average payload utilization, allocative utilization,

speed factor, or some combination of these factors. As expected, this shows that the

EEOI is highly influenced by how a ship is commercially operated, and only

partially influenced by the technical efficiency of the ship.

A.2 Data collection processing by company

Figure 3 provides an overview of all scripts developed in order to process the data

supplied by RBSA.

70

Figure 3 Matlab scripts structure

Company 1 (Bocimar) – script name: bocimar_rawsp

Calculation of main engine and auxiliary engine fuel consumptions ware obtained

using the formulas:

𝐹𝐶𝑚𝑒 = 𝐹𝐶𝑚𝑒_𝑑𝑎𝑦 ∗ 𝐴𝑇𝑑𝑎𝑦𝑠

𝐹𝐶𝑎𝑥 = 𝐹𝐶𝑎𝑥_𝑑𝑎𝑦 ∗ 𝐴𝑇𝑑𝑎𝑦𝑠

𝐹𝐶𝑝 = 𝐹𝐶𝑝_𝑑𝑎𝑦 ∗ 𝐴𝑇𝑝_𝑑𝑎𝑦𝑠

where :

o AT are the actual time days per voyage

o ATp are the actual time days at port

o FCme is the fuel consumption ME_day per voyage

o FC ax is the fuel consumption Aux_day per voyage

o FCp is the fuel consumption at port

71

Fuel type of the main engine is assumed to be HFO, while fuel type of auxiliary

engine and fuel consumption at port is assumed type to be MDO.

One single variable “distance” has been created merging three parameters: miles

laden, miles ballast and sea miles. When one of the first two is missing then sea

miles is used.

Company 2 (Delphis) – script name: delphis_collection_data

Data for this company are taken from noon report and arrival and departure report. The

noon report contains per each record:

the position of the ship (latitude and longitude) in a specific date,

the average speed,

the next port of call,

the distance made in the last 24 hours,

the miles to go to next port,

the fuels quantity on board,

the main engine RPM.

The arrival and departure report contains:

the date,

the port name,

fuels quantity on board on arrival and on departure,

the draft data on arrival and departure,

the next port of call.

We combine the noon reports and the arrival and departure reports, and we use two

temporal consecutive records to create a sea passage record. An algorithm is used to

calculate per each sea passage the distance made, the fuels consumptions at sea and at

port. We don’t have times data (hrs at sea) which have been calculated using speed and

distance.

The algorithm calculates the parameters for the EEOI calculation considering three

different cases:

1. ship at port and next signal at sea

2. ship at sea and next signal at port

3. case ship at sea and next signal at sea

Company 3 (Euronav) – script name: euronav_collection_data

Data are taken from the aggregated spreadsheet file provided by RBSA. This spreadsheet

doesn’t contain data for time at port, and fuel consumed in port. However time port in

72

hours is estimated using the dates of start and end of the sea passage assuming that each

sea passage finishes at port.

𝐻𝑟𝑠𝑝 = 𝑆𝑆𝑃(𝑡+1) − 𝐸𝑆𝑃(𝑡)

Where:

𝐻𝑟𝑠𝑝 are the hours spent at port

𝑆𝑆𝑃(𝑡+1) is the date of start of sea passage including time ('dd/mm/yyyy HH:MM') in the

observation (t+1)

𝐸𝑆𝑃(𝑡) is the date of start of sea passage including time in the observation (t)

We note that for these ships we still miss information about the fuel consumption at port.

Company 4 (Exmar) – script name: exmar_collection_data2

Data are taken from the aggregated spreadsheet file provided by RBSA.

Each row in this spreadsheet contains data per sea passage and data in adverse weather in

separated columns. So, data in adverse weather has been aggregated in each sea passage

and taken in account in the calculation of the indices.

Company 5 (Sea Tanker) – script name: seatanker_collection_data2

Data are taken from arrival and departure reports. This type of report contains per each

record:

port name,

arrival date and time,

departure date and time,

fuels quantities on board on arrival and on departure times,

cargo mass on board,

speed

distance

Two consecutive records are taken in account to create a sea passage record. The ESP is

calculated as difference between time of arrival and times of departure of the two records.

Similarly fuel consumptions at sea and at port is derived from fuels data at arrival and

departure date of the two records.

73

Cargo, speed and distance of the sea passage is assumed to be equal to the ones reported

in the latest record. Only when cargo is empty the sea passage is considered ballast.

A2. Outliners

The check of the consistency between the triple speed, distance and hours doesn’t

exclude outliners, which we define as “unrealistic data” (e.g negative values, very high

value). The Figure 4 shows the distribution of these variables and their correlations.

Figure 4 Distribution of speed, distance and time before excluding outliners

Then we remove the rows with all values that meet the following criteria:

• Speed >30 or <0 nmiles/hrs

• Distance >5000 or <0 miles

• Hours >5000 or <0 hr

Figure 5 shows the distribution after excluding outliners.

74

Figure 5 Distribution of speed, distance and time after excluding outliners

Similarly it has been applied for cargo. Every time cargo is bigger than the dwt, we

remove the entire row.

Slightly different it has been applied for the variables: hours at port, fuel consumptions at

ports, fuel consumptions at sea, and cargo. For these variables we set to NaN instead of

removing the entire row. Figure 6, Figure 7, Figure 8, and Figure 9 provide the

distribution of these variables before and after excluding the outliners. The criteria to

select the outliners for these variables are:

• Hors at ports >1000 or <0

• HFO and LSO at port >50 or <0

• MDO and MGO at port >100 or <0

• HFO at sea >10000 or <0

• LSO at sea >2000 or <0

• MDO and MGO at sea >100 or <0

75

Figure 6 Distribution of time and fuel consumptions at port before excluding outliners

Figure 7 Distribution of time and fuel consumptions at port after excluding outliners

76

Figure 8 Distribution of fuel consumptions at sea before excluding outliners

Figure 9 Distribution of fuel consumptions at sea after excluding outliners

A3. Excluded values.

In this step we generate “out” variables that are taken out from the calculation of the

EEOI. Particularly we take in account 4 cases:

1. If fuel consumption per laden voyage are known, however distance for these

voyages are missing and speed is different from zero. In this case for not

overestimating the numerator of EEOI formula we take out the fuel

77

consumptions and all the variables by setting NaN and creating the variables

“out”.

2. Fuel consumptions per laden voyages are missing, however distance for these

voyages are known and speed is different from zero. In this case for not

overestimating the denominator of EEOI formula we take out the distance and

all the variables by setting NaN and creating the variables “out”.

3. Fuel consumptions per laden voyages are known, however cargo for these

voyages are missing. In this case for not overestimating the numerator of

EEOI formula we take out the fuel consumptions and all the variables by

setting NaN and creating the variables “out”. In this case containers are

excluded.

4. Fuel consumptions per laden voyages are missing, however cargo for these

voyages are known. In this case for not overestimating the denominator of

EEOI formula we take out the cargo and all the variables by setting NaN and

creating the variables “out”. In this case containers are excluded.

Table 4 shows the number of ships in the sample for which we are able to calculate

the EEOI per year.

78

A3. Population Distribution Statistics

The population statistics (with a log normal y-axis scale and kernel distribution fit) for

categories of ship types is also calculated and represented in Figures 7-10.

Figure 10 Population statistics for Bulk carrier

Figure 11 Populations statistics for liquefied petroleum gas

79

Figure 12 Population statistics for containers

Figure 13 Population statistics for oil tankers

Understanding the Energy Efficiency Operational Indicator...2 Understanding the Energy Efficiency...

Documents

Transcript of Understanding the Energy Efficiency Operational Indicator...2 Understanding the Energy Efficiency...